On 21 Mar 2008, at 16:01, James Carlson wrote: > David Edmondson writes: >> >> On 20 Mar 2008, at 22:54, Peter Memishian wrote: >>> >>>> In the VNIC case, the failure of the underlying physical NIC >>>> (resulting in the link being 'down') doesn't stop the different >>>> VNICs >>>> from communicating with each other. As a consequence we decided not >>>> to >>>> propagate the link state. If there is just one VNIC then perhaps it >>>> would make sense to do so. >>> >>> Another alternative would be to provide another mechanism in.mpathd >>> could >>> use to conclude that the VNIC can no longer access other nodes >>> outside >>> the box. >> >> Doesn't probing already serve exactly that purpose? > > Slowly. Link up/down notification is typically very fast.
Slowly is better than nothing :-) Solving this 'properly' requires teaching in.mpathd about layer 2 reachability for arbitrary peers and then plugging in the various underlying mechanisms (VNICs, LDoms direct interdomain connections, ...). After that we'd have to somehow transform the layer 2 reachability into layer 3 reachability - maybe by fiddling in the forwarding table? It sounds like lots of work and, whilst I'd be happy to do it, I don't have time right now. _______________________________________________ networking-discuss mailing list [email protected]
